Confirmed neonatal bacterial infections occur in 1-5 per 1000 live births. However, in the US about 10-20 percent of newborns have "sepsis work-ups" done, and 4-10 percent receives systemic antibiotics. Textbooks recommend antibiotic therapy for infants with definite signs of sepsis, meningitis, shock, or respiratory failure. They do not provide evidence-based guidance for the evaluation and management of two common groups of term or near-term newborns: 1) infants with maternal risk factors for bacterial infection (e.g., chorioamnionitis) with no symptoms or who have presentations that are considered equivocal, and 2) infants with respiratory distress, which occurs in approximately 2-3 percent of term and near-term infants. Clinicians evaluating term or near term infants must make 3 decisions: 1) whether to obtain laboratory tests; 2) whether to treat with antibiotics, and 3) whether to transfer the infant to a tertiary care center. We propose to improve the evaluation and management of newborns >= 34 weeks gestation at risk for bacterial infection and/or critical illness by developing an evidence-based approach to estimating probabilities relevant to each of these decisions. We will integrate gestational age-specific prior probabilities with likelihood ratios for maternal risk factors, clinical signs, and age-specific laboratory results. To achieve this goal, we have these Specific Aims: 1) To perform a nested case-control study to quantify maternal and infant clinical risk factors for early onset bacterial infection; 2) To perform a retrospective cross-sectional study to estimate likelihood ratios for early onset bacterial infection for components of the complete blood count (CBC, the most common diagnostic test in this setting), using data from more than 40,000 CBCs and blood cultures from 14 hospitals; and (3) To perform a nested case control study to develop a quantitative model to estimate the probability of newborns >= 34 weeks gestation developing a critical illness (defined by life-threatening arterial blood gas results) based on clinical findings and the results of laboratory tests, including arterial blood gases. To achieve these aims, we will analyze paper and electronic records from 340,000 newborns >= 34 weeks gestation born from 1998 to 2005 at 14 hospitals in Northern California and Boston. Our project builds on considerable development work conducted by investigators at Kaiser Permanente's Division of Research, the University of California, San Francisco, and the Harvard Newborn Medicine Program. [unreadable] [unreadable] [unreadable]